Adaptive observation method combined with linear prediction model
By using an adaptive observation and quadratic programming linear model method, the uncertainty problem of control law in fuel cell stack electric vehicle system is solved, achieving efficient and reliable control setpoint determination and system adaptation, while reducing computational complexity and cost.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SCHAEFFLER TECHNOLOGIES AG & CO KG
- Filing Date
- 2024-11-13
- Publication Date
- 2026-04-10
AI Technical Summary
The modeling uncertainty of control laws in traditional industrial systems leads to unreliable control, especially in fuel cell stack electric vehicle systems, and existing algorithms are difficult to apply or computationally complex and expensive.
An adaptive observation method is adopted, which determines slowly varying parameters by programming linear differential equations, performs adaptive observation and quadratic programming of the optimal control setpoint, and uses a linear model for system adaptation and parameter identification.
It enables efficient and reliable control setpoint determination in fuel cell stack systems, adapts to system aging and changes, and reduces computational complexity and cost.
Smart Images
Figure CN121844264A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial control systems, and more specifically to the automatic optimization of the control laws of such systems.
[0002] More specifically, the present invention relates to a method for determining at least one control setpoint, and a circuit board in which the method of the present invention is implemented. Background Technology
[0003] Traditionally, control laws for industrial systems are formulated during the modeling phase, prior to their implementation within the system control unit. These control laws are based on models with inherent modeling uncertainties, leading to unreliable control for some applications, such as when controlling fuel cell stack electric vehicle systems.
[0004] Building a reliable model requires taking into account any evolution of the system over time, which is particularly relevant to aging effects. To account for these evolutions, a known practice is to determine empirical recalibrations by conducting experiments on real-world systems on a testbed. However, these recalibrations can be time-consuming, and the results of the experiments may not be representative of all systems in a given series.
[0005] To overcome these drawbacks, existing solutions integrate an algorithm into the system that adapts the control law to the system's actual state. However, often the control law is based on a physical model represented by nonlinear equations, making it difficult or even impossible to apply techniques that allow for parameter identification and adaptation. Furthermore, higher computational power is required to solve the nonlinear equations, meaning particularly complex and expensive computers are needed. Summary of the Invention
[0006] This invention proposes a method for adaptively observing a system at time t, the system comprising: - A set of input quantities, including the control setpoint u t And it is modeled in the form of the following vector: U t ; - A set of outputs, modeled as a vector: Y t ;as well as - A set of quantities representing the state of the system, modeled as a vector: x t ; The method is implemented in a computational unit and includes a first step E1 of programming a set of linear differential equations related to these inputs, these system state variables, and these outputs, which are in the following format: [Mathematical Expression 13] - where θ t = { , … , } are slowly varying parameters, which change more slowly than U over time. t Y t and x t ;and - Where A(θ) t ), B(θ) t ), C(θ) t ) and D(θ t ) is a collection of these slowly varying parameters { , … , A matrix consisting of 1, -1, or 0; - Step E2, which involves minimizing the measured value of the output (Y). t ) mes. These output quantities Y, calculated using mathematical equation 13 t The difference e between the values determines the slowly varying parameter θ. t = { , … , }: [Mathematical Expression 14] The method includes: - Step E3, which is performed after or concurrently with step E2, is the adaptive observation step. This adaptive observation step includes numerically solving equation 13 by considering the minimum difference e at time t to determine the quantity x of the system state at time t. t Perform numerical calculations.
[0007] In step E2 or step E3, these slowly varying parameters θ t = { , … , }, where the system state quantity x t It is obtained numerically by solving equation 13 iteratively and minimizing the difference e several times.
[0008] The present invention also relates to a method for determining at least one control setpoint u of a system at time t. t The method includes: - Steps E1, E2, and E3 of the adaptive observation method according to the present invention; - Step E4 involves determining the optimal control setpoint within a time window [T0, T0+NΔT] containing N samples using quadratic programming. This determination process utilizes the following quadratic criteria to be optimized: [Mathematical Expression 15] - where x 理想 It is a vector representing the target state of the system; - in, It is a vector The transpose of the vector; and - Where Q, R, and K are the weighting matrices of the quadratic criterion to be optimized; - Where N is a natural integer; Then, the control setpoint u is obtained for N samples from the current time Tp in the future. t For time T p+i = T p Each sample of +iΔT represents the setpoint as a function of the following terms: A(θ Tp+1 ), B(θ) Tp+1 ), C(θ) Tp+1 ) and D(θ Tp+1 ), x Tp+1 Q, R, and K.
[0009] This method may include when these slowly varying parameters θ t = { , … , When any one of the elements in} changes, a new iteration is performed on step E4, which uses the quadratic programming method to determine the optimal control setpoint.
[0010] The method may include step E5, which involves sending the next time step T to the system control unit after each new iteration of step E4, which determines the optimal control setpoint. p+1 The optimal control setpoint.
[0011] This method can be implemented in a fuel cell stack system, which includes: - Battery; - Fuel cell stacks; and - Electric motor; Among them, [Mathematical Expression 16] ,in: - I bat It is the ampere number of the current in the battery; - I FC It is the number of amperes of the current in the fuel cell stack; - f(I FC ) is V FC The calculations involving I FC The function; Among them, [Mathematical Expression 17] ,in: - V bat It is the voltage in the battery; - V FC It is the voltage in the fuel cell stack; - m H2 This refers to the amount of hydrogen consumed in the fuel cell stack. Among them, [Mathematical Expression 18] ,in: - SoC is the state of charge of the battery; - m H2 This is the amount of hydrogen consumed in the fuel cell stack; - V 损耗 It is the voltage associated with electrical losses in the system, which characterizes the aging state of the fuel cell stack; The linear mathematical equation 13 for online adaptation applies to the aging V of the fuel cell stack. 损耗 The dynamic characteristics are modeled.
[0012] Matrices A, B, C, and D can be in the following format: [Mathematical Expression 19] [Mathematical Expression 20] [Mathematical Expression 21] [Mathematical Expression 22] This method can be implemented in systems susceptible to fouling effects or friction.
[0013] The present invention also relates to a circuit board in which the method according to the invention is implemented. Attached Figure Description
[0014] Further features and advantages of the invention will become apparent from the following detailed description, which will be understood with reference to the accompanying drawings, in which: [ Figure 1 [Illustration] is a graph showing a set of system control setpoint values changing over time, obtained by a method according to an embodiment of the present invention; [ Figure 2 [ ] is a flowchart illustrating the sequence of steps in a method according to an embodiment of the present invention; [ Figure 3 ] is shown Figure 2 Flowcharts for steps E2 and E3 of the method; [ Figure 4] is shown Figure 2 A flowchart showing the sequence of steps E2 to E5 in the method; [ Figure 5 [Illustrated representation of the operation of components in a fuel cell stack electric vehicle system, the method of the present invention can be applied to this system;] [ Figure 6 [This is an illustrative illustration of applying the method of the present invention to...] Figure 5 The flowchart of the system. Detailed Implementation
[0015] In the following description, the same, similar or analogous elements are indicated by the same reference numerals.
[0016] Figure 1 This shows the system's control setpoint u. t A graph showing a set of values changing over time. These values are obtained using the method of this invention, thereby allowing the use of predictive models to control the system. These values are calculated for each time increment k, where, in the example, k ranges from 0 to N.
[0017] The method of this invention allows for the calculation of the optimal control setpoint u for each future time increment within a given time window. k This calculation is performed during step E4 of the method, which will be detailed below. The calculation implements the direct quadratic programming method, which is known in itself.
[0018] Each iteration of step E4 allows for the calculation of a set of optimal control setpoint values u for future time increments within a given time window. k The set of values obtained after one iteration in step E4 constitutes the control trajectory. Figure 1 The graph includes multiple control trajectories generated by each iteration of step E4. Each control trajectory is plotted with a specific relative dark line.
[0019] Only when the system parameter θ k The new control trajectory is calculated only when the condition changes, i.e., a new iteration is performed on step E4. θ k This represents a parameter that evolves, for example, under the aging effect of the system. In this example, the parameter θ changes between the increments k = 2 and k = 3. k The value remains unchanged, i.e., θ3 = θ2. Therefore, the control trajectory was not recalculated at increment k = 3, but was recalculated at other increments.
[0020] The dashed line represents the actual control setpoint u sent to the system control unit. k For each increment k, the actual control setpoint u sent to the system control unit. k This corresponds to the last control setpoint value calculated for the increment k under discussion.
[0021] Figure 1 The method of the present invention is shown to enable the use of predictive models to control the system by determining an optimal control setpoint that adapts to the actual behavior of the system.
[0022] Figure 2 The steps of a method according to an embodiment of the present invention are illustrated in the form of a flowchart.
[0023] The method of the present invention is implemented in a computing unit of a system including, for example, a general-purpose processor, a microprocessor, or even electronic circuitry. The computing unit may form part of a system control unit, or may include means for wired or wireless communication with the system control unit. Furthermore, the computing unit in which the method is implemented includes a memory capable of storing values of certain quantities used by the algorithm of the method.
[0024] The method includes the following steps: - The first step E1 in programming a set of linear differential equations related to the inputs, system state variables, and outputs; - The second step in determining the slowly varying parameters, E2; - The third step of adaptive observation, E3, includes numerical calculation of the system's state at time t; - Step E4, which uses quadratic programming to determine the optimal control setpoint; - Step E5: Send the optimal control setpoint for the next moment to the system.
[0025] The first step, E1, corresponds to implementing the system behavior model in the system computing unit.
[0026] The first step, E1, involves programming a set of linear differential equations within the system's computational unit. These linear differential equations establish the relationships between input variables, system state variables, and output variables. These equations are in the following format: [Mathematical Expression 1] - where θ t = { , … , } are slowly varying parameters, which change more slowly than U over time. t Y t and x t ;and - Where A(θ) t ), B(θ) t ), C(θ) t ) and D(θ t ) is a collection of these slowly varying parameters { , … , A matrix consisting of 1, -1, or 0.
[0027] Slowly varying parameter θ t = { , … , These are parameters that do not necessarily have physical meaning, and these parameters change slowly over time, especially due to the effects of aging and wear of system components.
[0028] The second step, E2, corresponds to the identification of the system parameters. The second step, E2, includes determining the slowly varying parameter θ. t = { , … , }
[0029] Step E2 is based on the following reasoning: This reasoning includes considering the measured value of the output quantity (Y). t ) mes Theoretically, it must be equal to the output quantity Y calculated using Equation 1. t The value of is obtained by incorporating the measured value into the equations and solving the system of equations. The unknowns of the system can then be deduced.
[0030] In practice, the slowly varying parameter θ t This is obtained by minimizing the difference e between the following terms: - Measured output value (Y) t ) mes ;as well as - Output quantity Y calculated using Equation 1 t The value of . The difference e is expressed in the following format: [Mathematical Expression 2] .
[0031] Obtaining zero difference e indicates that the identification algorithm converges well to the parameter θ. t The index of the correct value. This convergence is subject to the input quantity u. t The influence of the conditions on the persistence of the excitation. This is achieved by adjusting the optimal control setpoint u. k This is ensured by adding a signal with very low amplitude and abundant frequency. The low amplitude results in negligible impact on the cost function J, which is intended to be optimized.
[0032] The third step, E3, corresponds to adaptive observation, which involves estimating the system state based on a model that is continuously adapted throughout the system's lifecycle. This involves updating the model based on the evolution of system behavior, particularly as the system ages.
[0033] The third step, E3, is performed after step E2, such that during step E3, the slowly varying parameter θ...t This can be used to calculate values characterizing the system state. Step E3 involves considering the minimum difference e at time t and the slowly varying parameter θ obtained in step E2. t Numerical solutions are used to obtain the state quantity x of the system from equation 1. t Perform numerical calculations.
[0034] As a variant, steps E2 and E3 are performed simultaneously via a unified solution of these equations.
[0035] The fourth step, E4, corresponds to representing optimal control.
[0036] Step E4 involves determining the optimal control setpoint u using quadratic programming. t Applying quadratic programming within a time window [T0, T0+NΔT] containing N samples, the application process uses the following quadratic criterion to be optimized: [Mathematical Expression 3] - where x 理想 It is a vector representing the target state of the system; - in, It is a vector The transpose of the vector; and - Where Q, R, and K are the weighting matrices of the quadratic criterion to be optimized; - Where N is a natural integer.
[0037] Each term of the quadratic criterion contains the product of a vector and its transpose, hence it is called a "quadratic form".
[0038] The direct quadratic programming method is well-known in the literature and is primarily used to obtain the optimal control setpoint for a linear model within a time window containing N observations. Therefore, the details of the calculations performed using the quadratic programming method are not included in this specification.
[0039] By calculating the minimum value of the quadratic criterion J, the control setpoint u is obtained for N future samples starting from the current time Tp. t For time T p+i = T p For each sample of +iΔT (i is between 0 and N), the setpoint is represented as a function of the following terms: A(θ Tp+1 ), B(θ) Tp+1 ), C(θ) Tp+1 ) and D(θ Tp+1 ), x Tp+1 The terms Q, R, and K can be represented numerically, especially by means of θ, which makes it possible to obtain θ. Tp+1Step E2 and making x obtainable in numerical form Tp+1 Step E3.
[0040] Step E5 corresponds to using a predictive model to control the system.
[0041] Step E5 includes sending the next time step T to the system control unit. p+1 Optimal control setpoint u Tp+1 The optimal control setpoint is sent after each new iteration of step E4, which determines the optimal control setpoint.
[0042] This strategy of using predictive models to control the system is reliable because it continuously adapts to the actual operation of the system, especially taking into account how the system operation evolves with aging.
[0043] Steps E2 through E5 are executed during system operation. These steps can be embedded, i.e., implemented in computing and control units integrated into the system.
[0044] Figure 3 Steps E2 and E3 are illustrated in flowchart form. Step E2, which identifies system parameters, and step E3, which involves adaptive observation, are two closely related steps. In steps E2 and E3, the state estimator x... k+1 and parameter estimator θ k+1 Generate a predicted value for the next time increment k+1 within each time increment k.
[0045] Figure 3 This shows that, in order to estimate the next time increment k+1, the adaptive observer requires the following as input data: - Input quantity U sent to the control unit or measured k The value; - Measured output value (Y) k ) mes. With output quantity Y k The difference e between the calculated values, which is provided, for example, by comparator 12; - The quantity x of the system state within the current time increment k k and parameter θ k The values are stored in memory 10.
[0046] The estimated system state x for the next time step k+1 k+1 and parameter θ k+1 These are the output data from steps E2 and E3, which will then be used in step E4 to update the optimal control setpoint u. k+1 .
[0047] In step E2, by iteratively solving Equation 1 within the time window [T0, T0+NΔT] and minimizing the difference e multiple times, these slowly varying parameters θ can be obtained numerically. t = { , … , In fact, recursive computation allows asymptotic convergence to θ. t The correct value.
[0048] Similarly, in step E3, by iteratively solving Equation 1 within the time window [T0, T0+NΔT] and minimizing the difference e multiple times, the system state quantity x can be obtained numerically. t The value of .
[0049] Figure 4 The sequence of steps E2 to E5 is illustrated in flowchart form. The predictive model-based control strategy behind this sequence of steps E2 to E5 aims to establish a relationship between the updated model and the optimal control setpoint in order to obtain a reliable optimizer.
[0050] Figure 4 The hierarchy of each step is shown in a three-column table. These steps are divided into three levels, as follows: - Steps E2 and E3 are at the level of adaptive observation, that is, the level at which the system state is determined based on the updated measured values; - Step E4 is located at a level in the optimization algorithm, that is, representing the level of optimal control; and - Step E5 is located at the level of using predictive models to control the system, i.e., at the control level.
[0051] Following steps E2 and E3, a logic test 14 is performed to verify the completion of step E4, which determines the optimal control setpoint. This logic test has the following conditions: θ k+1 ≠ θ k In other words, when the slowly varying parameter θ t = { , … , When one of the conditions θ changes, step E4 is triggered. Upon completion of step E4, step E5, which sends the optimal control setpoint to the system, is automatically triggered. Therefore, if condition θ is not met... k+1 ≠ θ k Then, the setpoint u corresponding to the intended transmission at increment k+1 is... k+1 The command will not be relative to an existing setpoint u in memory. k+1 The value is modified, and the update of the control is disabled, as shown in step E6.
[0052] In fact, if the slowly varying parameter θ tTo keep the control setpoint constant, updates to the optimal control setpoint are disabled to avoid unnecessary calculations, since the control setpoint u is calculated for N samples starting from time Tp. t The value remains constant. Therefore, the control unit at each time T p+i = T p +iΔT will take into account the last control setpoint u calculated for that moment. Tp+i .
[0053] If from time T P Since then, no optimal control setpoint u has been executed. t The update can be specified in the time window T. P When +NΔT ends, a new iteration is performed on step E5.
[0054] The predictive model-based control strategy uses an internal model that is continuously adapted throughout the system's lifecycle by means of an adaptive observation method.
[0055] Figure 5 The operation of components of a fuel cell stack electric vehicle system 16 is illustrated schematically, and the method of the present invention can be applied to this system. In the example, the vehicle is a car or truck. As a variation, the vehicle can be any other type of transportation, such as a boat or an airplane.
[0056] System 16 includes: - Junction box 18, which allows for centralized management of electrical connections between different electrical devices; - Battery 20; - Fuel cell stack 22; - Auxiliary components 24 of the fuel cell stack; and - Electric motor 26.
[0057] Arrows indicate the exchange of power between components.
[0058] Battery 20 is capable of supplying or storing electrical energy, as indicated by a double-headed arrow pointing to battery 20 to symbolize that the battery can receive and store electrical energy in storage mode, and pointing to junction box 18 to symbolize that battery 20 can supply electrical energy in power supply mode.
[0059] The fuel cell stack 22 allows for the supply of electrical energy by oxidizing a reducing fuel (such as hydrogen) at one electrode and reducing an oxidant (such as oxygen from the air) at another electrode. The fuel cell stack 22 transmits the generated electrical energy to the junction box 18.
[0060] The auxiliary components 24 of the fuel cell stack include, for example, a compressor and a DC-DC converter, which is a power converter that transforms a DC source from one specified voltage level to another different voltage level. The auxiliary components 24 consume a portion of the electrical energy.
[0061] The electric motor 26 consumes electrical energy to supply motor torque, but when it operates as a generator, such as during deceleration, the electric motor can also supply electrical energy.
[0062] The power balance among the 18 terminals of the junction box is represented by the following equation.
[0063] [Mathematical Expression 4] in: - Pw FC The power is supplied by the fuel cell stack 22; - Pw bat This is the power supplied by battery 20 (the sign is positive in power supply mode and negative in storage mode). - Pw req This is the power supplied to the electric motor 26 (the sign is positive if the electric motor 26 consumes energy, and negative when the electric motor 26 operates as a generator). - Pw aux The power consumed by auxiliary component 24.
[0064] The system includes: - A set of input quantities, which are modeled in the form of a vector: U t ; - A set of outputs, modeled as a vector: Y t ;as well as - A set of quantities representing the state of the system, modeled as a vector: x t .
[0065] Inputs, outputs, and quantities representing system state are quantities that change with time t. In the following expressions, the subscript "t" indicating time dependence is sometimes omitted to simplify the text.
[0066] Input and output quantities are measurable quantities, or quantities that can be derived from known information about the system.
[0067] Input quantities include: - The ampere current supply of the fuel cell stack FC The ampere number is the variable to be controlled and regulated, and the optimal control setpoint u needs to be calculated for this variable. t ;as well as - Power Pw to be delivered to the electric motor req This power is not a controllable quantity, but it is considered known given the vehicle's known range.
[0068] The ampere number I of the supply current of the fuel cell stack FC This determines the operating condition of the fuel cell stack. Therefore, the ampere number I of the power supply current of the fuel cell stack... FC The larger the value, the greater the amount of electrical energy generated by the fuel cell stack, and therefore the greater the instantaneous power Pw delivered by the fuel cell stack. FC The larger the current, the greater the current. Therefore, the ampere number I of the fuel cell stack's power supply current. FC The larger the value, the more power Pw is provided. FC The more hydrogen is consumed.
[0069] The power Pw to be delivered to the electric motor req Through the power balance and power Pw between the terminals of junction box 18 as described above FC ,Pw bat and Pw aux Related. Therefore, due to Pw req It can be derived from known information about the vehicle's journey, therefore Pw can be... FC ,Pw bat and Pw aux The quantities involved in the calculation (i.e., I) bat and I FC ) is introduced into the input variables.
[0070] Figure 6 This illustrates the application of the method of the present invention. Figure 5 The flowchart of system 16.
[0071] The input is modeled by a vector U, as shown below: [Mathematical Expression 5] ,in: - I bat It is the ampere number of the current in the battery; - I FC It is the number of amperes of the current in the fuel cell stack; - f(I FC ) is V FC The calculations involving I FC A function of . More specifically, it is when the fuel cell stack system is in its brand new state (no aging or V aging). 损耗 Voltage-current characteristics when = 0).
[0072] Output quantities include: - Voltage V in the battery bat ; - Voltage V in fuel cell stack FC ; - Hydrogen consumption m in the fuel cell stack H2 .
[0073] The output is modeled by the vector Y, as shown below: [Mathematical Expression 6] ,in: - V bat It is the voltage in the battery; - V FC It is the voltage in the fuel cell stack; - m H2 This is the amount of hydrogen consumed in the fuel cell stack.
[0074] Quantities representing system state include: - Battery state of charge (SoC); - Hydrogen consumption m in the fuel cell stack H2 ; - Voltage V related to electrical losses in the system 损耗 This voltage characterizes the aging state of the fuel cell stack.
[0075] The quantity representing the system state is modeled by the vector x, as shown below: [Mathematical Expression 7] ,in: - SoC is the state of charge of the battery; - m H2 This is the amount of hydrogen consumed in the fuel cell stack; - V 损耗 It is the voltage related to electrical losses in the system.
[0076] exist Figure 5 and Figure 6 In the system, Equation 1 is expressed as follows: [Mathematical Expression 8] Matrices A, B, C, and D are in the following format: [Mathematical Expression 9] [Mathematical Expression 10] [Mathematical Expression 11] [Mathematical Expression 12] The slowly varying parameters are related to the aging of the battery 20 and the fuel cell stack 22. In fact, the behavior of the battery 20 and the fuel cell stack 22 changes with their use.
[0077] Other equations allow for the representation of the operational constraints of system 16. Therefore, the relationship between current and voltage within battery 20 is represented. Similarly, the relationship between current and voltage within fuel cell stack 22 is represented.
[0078] In the quadratic criterion J to be optimized (i.e. minimized), the first term (containing x) t ) reflects the aging of system 16, and the second item (containing U) t This reflects the consumption, especially the consumption of hydrogen. Therefore, minimizing the quadratic criterion J truly achieves the optimization, i.e., minimization, of both the aging of the system and its consumption.
[0079] The algorithm is used to calculate the ampere number I of the current passing through the fuel cell stack 22. FC To calculate the energy required for fuel cell stack 22 to achieve several objectives, namely: - Minimize the amount of hydrogen consumed over a known distance; - Minimize system aging, and in particular minimize the aging of fuel cell stack 22 and cell 20.
[0080] The algorithm is implemented on the circuit board that controls the power flow at junction box 18.
[0081] This invention can be used in other systems. For example, it can be used to design optimal control for systems such as: - Cooling loop systems based on heat exchangers to improve reliability against heat exchanger fouling effects; - Motor, to improve the reliability of friction.
[0082] In both examples, the dynamic model will be modified and adapted to the inputs, outputs, and system state variables involved in the system under discussion.
[0083] The present invention has many advantages, which are described below.
[0084] The method for determining control setpoints provided by this invention is reliable and accurate because it is adaptive and based on actual measurements taken within the system.
[0085] This method is based on finding a mathematical model that relates to the system’s inputs and outputs in a linear form, so as to obtain a model that is accurate enough to allow the implementation of algorithms to adapt and identify the parameters.
[0086] The control setpoint is expressed as a function of the following: A(θ Tp+1 ), B(θ) Tp+1 ), C(θ) Tp+1 ) and D(θ Tp+1 ), x Tp+1 The inclusion of Q, R, and K allows for the relatively simple implementation of the deterministic algorithm within the computational unit. In particular, the measurement steps are executed quite rapidly, and the implementation of the solution algorithm does not have high computational time requirements. Therefore, in principle, real-time implementation of this method is not a problem.
[0087] Therefore, the method of the present invention can be implemented by a computing unit and a control unit comprising a relatively simple and inexpensive processor.
[0088] This invention provides a reliable embedded solution for optimizing system energy consumption and aging, particularly for fuel cell stack electric vehicles.
Claims
1. A method for adaptively observing a fuel cell stack system at time t, the system comprising: - Battery (20); - Fuel cell stack (22); as well as - Electric motor (26); - A set of inputs, including the control setpoint ut, and modeled as a vector: Ut; - A set of outputs, modeled as a vector: Yt; and - A set of quantities representing the state of the system, modeled as a vector: x t ; The method is implemented in a computational unit and includes a first step E1 of programming a set of linear differential equations related to these inputs, these system state variables, and these outputs, which are in the following format: [Mathematical Expression 13] - where θ t = { , … , } are slowly varying parameters, which change more slowly than U over time. t Y t and x t ;and - Where A(θ) t ), B(θ) t ), C(θ) t ) and D(θ t ) is a collection of these slowly varying parameters { , … , A matrix consisting of 1, -1, or 0; - Step E2, which involves minimizing the measured values of these output quantities (Y) t ) mes. With the use of equations The calculated output quantity Y t The difference e between the values determines the slowly varying parameter θ. t = { , … , }: [Mathematical Expression 14] The method is characterized in that it includes: - Step E3, which is performed after or concurrently with step E2, is the adaptive observation step. This adaptive observation step includes numerically solving the following equation to determine the quantity x of the system state at time t by considering the minimum difference e at time t. t Perform numerical calculations: and Among them, [Mathematical Expression 16] ,in: - I bat It is the number of amperes of the current in the battery (20); - I FC It is the number of amperes of the current in the fuel cell stack (22); - f(I FC ) is V FC The calculations involving I FC The function; Among them, [Mathematical Expression 17] ,in: - V bat It is the voltage in the battery (20); - V FC It is the voltage in the fuel cell stack (22); - m H2 This is the amount of hydrogen consumed in the fuel cell stack (22); Among them, [Mathematical Expression 18] ,in: - SoC is the state of charge of the battery (20); - m H2 This is the amount of hydrogen consumed in the fuel cell stack (22); - V 损耗 It is the voltage associated with electrical losses in the system, which characterizes the aging state of the fuel cell stack (22); Linear equations that enable online adaptation Aging V of the fuel cell stack 损耗 The dynamic characteristics are modeled.
2. The method as described in the preceding claim, characterized in that, In step E2 or step E3, these slowly varying parameters θ t = { , … , }, where the system state quantity x t The following equation was solved iteratively several times, and the difference e was minimized several times, and the result was obtained numerically. 。 3. A method for determining at least one control setpoint u of a system at time t. t The method is characterized by, The method includes: - Steps E1, E2, and E3 of the adaptive observation method as described in any one of claims 1 and 2; - Step E4 involves determining the optimal control setpoint within a time window [T0, T0+NΔT] containing N samples using quadratic programming. This determination process utilizes the following quadratic criteria to be optimized: [Mathematical Expression 15] - where x 理想 It is a vector representing the target state of the system; - in, It is a vector The transpose of the vector; and - Where Q, R, and K are the weighting matrices of the quadratic criterion to be optimized; - Where N is a natural integer; Then, the control setpoint u is obtained for N samples from the current time Tp in the future. t For time T p+i = T p Each sample of +iΔT represents the setpoint as a function of the following terms: A(θ Tp+1 )、B(θ Tp+1 )、C(θ Tp+1 ) and D(θ Tp+1 ), x Tp+1 , Q, R, and K.
4. The method as described in claim 3, characterized in that, The method includes when these slowly varying parameters θ t = { , … , When any one of the elements in} changes, a new iteration is performed on step E4, which uses the quadratic programming method to determine the optimal control setpoint.
5. The method as described in any one of claims 3 and 4, characterized in that, The method includes step E5, which involves sending the next time step T to the system control unit after each new iteration of step E4, which determines the optimal control setpoint. p+1 The optimal control setpoint.
6. The method as described in the preceding claim, characterized in that, Matrices A, B, C, and D are in the following format: [Mathematical Expression 19] [Mathematical Expression 20] [Mathematical Expression 21] [Mathematical Expression 22] 。 7. The method according to any one of claims 1 to 5, characterized in that, This method is implemented in systems susceptible to fouling effects or friction.
8. A circuit board wherein the method as described in any one of claims 1 to 7 is implemented.